Three conditions, all at once, or reach for something else

Complexity thinking has a narrow home. It earns its place when behaviour emerges from interaction, the rules change as things unfold, and the people looking are part of what they see. Where those three do not hold together, other forms of thinking do the job better, and this page says which.

Complexity

All three at once: complexity’s home ground

Emergence

The behaviour that matters comes from interactions among the parts and cannot be read off the parts themselves (Anderson, 1972; Holland, 1998). Knowing every actor’s intentions does not tell you the outcome.

Nobody at the school decided which subjects were harsh. Students compared ratings in the canteen, saw the same work rated Developing in science and Secure in English, and told each other. By the end of Term 1, Week 4, Secondary 2 had a ranking of subjects that no teacher had written. No single student’s intention explains it.

Adaptation

The agents, their connections and the rules that govern them change as the situation unfolds, so the structure itself is in motion (Holland, 1995; Arthur, 2015). A map drawn today describes a system that no longer exists tomorrow.

In Term 1, Week 6 a class parents’ chat group decided that without weighted assessments nobody knew where their children stood, and some families signed their children up for tuition centres that set practice papers. Part of Secondary 2 now had a second assessment system, outside the school and running on the old rules. Any map of how Secondary 2 was assessed, drawn in Week 5, was out of date by the next week.

Reflexivity

The analyst’s framing, measurement and action enter the situation and alter it (Stacey, 2001; von Foerster, 2003). Publishing the forecast changes the outcome. Setting the target changes the behaviour.

The school set a target of 95 per cent of books showing feedback in the agreed format, and checked it with a book check. Teachers started writing feedback for the book check, not the student, and humanities read its “inconsistent” rating as blame. By Term 2, Week 5 the dashboard showed 96 per cent and nobody in the school trusted it.

One at a time

Each condition alone already has a home

Two of the three, without the third, belong to a discipline that handles them well. Reach for that discipline first.

Each partial combination and the discipline it belongs to
Has	Lacks	Example	Already belongs to
Emergence	Adaptation and reflexivity	A traffic jam, or a weather system. The pattern comes from interaction, but the rules each car or air mass follows stay put, and the forecast does not change the weather.	Statistical physics and simulation
Adaptation	Reflexivity	An ecosystem. Species, relationships and niches change over time, but the ecologist studying the pond does not change it by writing about it.	Ecology and evolutionary biology
Reflexivity	Emergence	A negotiation between two parties. Each side’s reading of the other changes what it does, but the outcome follows from two sets of moves, not from a crowd of interactions.	Game theory and strategy

All three

Why the conjunction matters

Each condition alone is already covered

Taken one at a time, none of the three conditions needs a new discipline. Emergence on its own is the business of statistical physics and simulation. Adaptation on its own belongs to ecology and evolutionary biology. Reflexivity on its own is what game theory and strategy were built for.

Each of those disciplines works because it can hold the other two conditions still. A traffic simulation assumes the drivers’ rules do not change. An ecologist assumes the pond does not read the field notes. A game theorist assumes two players whose moves can be listed.

All three at once

When all three combine, as they do in markets, health systems, cities, geopolitics, information ecosystems and organisations in transformation, no single existing discipline is adequate. Nothing can be held still. The pattern comes from interaction, the rules shift while you watch, and your watching is one of the things shifting them.

That conjunction is complexity thinking’s home ground. It is also why the test is a test. A situation that meets one or two of the conditions has a better-equipped discipline waiting for it.

Why human systems almost always qualify

People interact, so patterns emerge. People learn and change their habits, so the rules move. And people notice when they are being measured, so the measure becomes part of the situation. That is why the field matters most in human systems, and why a school’s new assessment plan meets all three conditions within a fortnight of the first ratings going home.

What it excludes

Complicated is not complex

The test earns its keep by what it rules out. Using complexity methods on these is a category error.

A bridge

Complicated, not complex. It has many parts and needs real expertise, but the parts do not change their behaviour as the engineers study them. Calculation works, and it can be checked.

A payroll migration

A technical problem, not an adaptive challenge (Heifetz, 1994). Experts can solve it, and the people it affects do not have to change their values or habits for it to succeed.

The school timetable

Complicated. Hundreds of constraints, but they hold still long enough to be solved, and a good timetabler or a good piece of software can find an answer that anyone can check. Running safe-to-fail probes on it would waste a term.

Setting up the tracking system

Technical. Price, data protection, and whether ratings reach the school’s parent app are questions with answers. The semester shows where the line sits. Setting up the system was technical. Deciding what a five-point scale should mean, and what parents would read into it, was not.

The neighbours

Thirteen ways of thinking, and what each holds still

Every neighbouring way of thinking treats something as given. Complexity thinking is the one that does not.

Neighbouring ways of thinking compared
Way of thinking	Its question	Treats as given	Blind spot
Analytical / reductionist	What are the parts, and how do they work?	That wholes are explained by parts	Interaction effects and emergent behaviour
Critical thinking	Is this claim warranted by evidence and argument?	A stable object about which claims can be checked	Claims that change the object they describe
Systems thinking	How is the system structured, and where are the leverage points?	A structure that can be mapped and that holds still	Structural novelty, and the analyst inside
Design thinking	What do users need, and what could we prototype?	A bounded user and problem	Second-order and system-level effects
Abductive reasoning	What might explain this surprising observation?	That an explanation is there to be found	Explanations that shift as the system adapts
Lateral thinking	What other way of seeing this is there?	A cognitive pattern to escape	Why some reframings take hold and others do not
Strategic thinking	How do we reach our goals against rivals and constraints?	Stable goals and a knowable competitive field	Goals that emerge, and rivals who adapt to the strategy
Futures thinking	What futures are possible, probable and preferable?	That futures can be imagined in advance	The present’s dispositions, and futures that emerge from action
Integrative / paradox thinking	How can opposing models be held and transcended?	A small number of identifiable poles	Many interacting tensions across scales
Crisis thinking	What is breaking, how are breakdowns linked, and what generates them?	A norm from which crisis departs	Crisis as regime, not exception, and mood replacing mechanism
Metamodern thinking	How can we commit sincerely to narratives we know are partial?	That stance and sensibility are the levers	Mechanisms and evidence
Warm data / relational inquiry	What relationships across contexts are we missing?	That relational information is reachable through conversation	Scale and testability
Complexity thinking	What is this situation disposed to do, how will it respond to our acting on it, and how will we keep learning?	Very little. Structure, rules and the analyst’s position are all treated as variables	Precision, accountability and over-application (see the failure modes below)

Systems thinking

The nearest neighbour

What the others hold still

Read down the column of what each form treats as given and a pattern appears. Each of the others assumes that something holds still long enough to be known: the parts, the evidence, the structure, the user, the goal, the poles of a tension, the norm from which a crisis departs.

Complexity thinking declines that assumption. Its object is the situation’s own capacity to change, including in response to being thought about. So it cannot be one more technique in the toolbox. It is the form of thinking that asks when the other techniques’ assumptions hold.

Complement, not rival

None of this makes complexity thinking superior. It gives it a different job. Analytical thinking is still the tool for components, critical thinking for claims, systems thinking for structure, design thinking for user value, abduction for hypotheses and lateral thinking for reframing.

Complexity thinking supplies three things none of them does. A diagnostic: what kind of situation is this, and so which forms of thinking are valid? A dynamics layer: what will this become, and what will our action set in motion? And a stance: we are inside, our models are partial, and our commitments must be revisable.

The three closest neighbours

Three forms are most often confused with complexity thinking. Systems thinking maps structure. Complexity thinking studies how structure emerges and shifts.

Abductive reasoning generates hypotheses. Complexity thinking explains why, in adaptive systems, hypotheses have to be tested by acting rather than by analysis alone. Abduction is its engine and probing is its method (Dorst, 2011; Snowden & Boone, 2007).

Futures thinking imagines alternative futures. Complexity thinking insists that futures grow out of the dispositions of the present and the probes run now, so the most useful futures work is anchored in present propensities (Jullien, 2004; Sharpe et al., 2016).

Fixed structure

System dynamics represents a situation as stocks, flows and feedback loops, with the structure specified by the modeller. Behaviour then follows from the structure. In an adaptive system the structure itself changes: new agents, new connections and new rules appear. A model with a fixed structure cannot produce real novelty.

The boundary problem

A system has to be bounded to be modelled. In complex systems the world does not supply the boundary. The observer draws it (Churchman, 1968; Ulrich, 1983). Cilliers (2001) showed that such boundaries are often porous and multiple, and can sit inside and outside one another at once. Every boundary choice lets some stakeholders in and leaves others out.

The designer outside the system

Stacey (2001) argued that much systems thinking imagines a manager who stands outside the system, maps it and redesigns it. In human systems nobody stands there. The strategist is part of the pattern being changed, and everyone else reinterprets what they do. Of the six critiques, this one has the largest practical consequences.

Leverage-point optimism

Meadows’ (1999) leverage points are among the field’s most cited ideas, and Meadows herself warned that they are counterintuitive and often pushed the wrong way. Popular systems thinking tends to promise that a well-placed nudge will shift a system. Complexity says the same nudge can do nothing, do a great deal or do the opposite, depending on a state of the system nobody can know in advance.

Metaphor in place of model

A 2025 essay by Prairie Oyster argues that the language of interconnection is now “table stakes”. On this reading, systems thinking says what to look at, the whole and its connections, but not how to act under uncertainty, and so risks becoming a comforting story instead of a practical guide. Its remedy is complexity literacy: models over metaphors, pluralism over purity, iteration over nudging, and governance that takes power and incentives seriously.

Power and politics

Critical systems thinkers (Ulrich, 1983; Jackson, 2019) have long argued that system maps take the politics out. Whose purposes define the system, and whose knowledge counts in modelling it? Complexity science does not answer these either. It is strong on description and thin on values. But the questions get harder to ignore once every model is accepted as a partial, interested framing (Cilliers, 1998).

A difference of assumptions, not of tools

“Systems thinking” is at least four traditions: hard systems approaches, system dynamics, soft systems methodology and critical systems thinking. The critiques above mostly land on the first two. Soft and critical systems thinking have already absorbed much of the complexity critique.

The most defensible reading is that systems thinking and complexity thinking are not rival toolkits. They rest on different default assumptions. Systems thinking usually assumes that structure can be identified and that behaviour follows from it, so the way to intervene is to redesign the structure. Complexity thinking assumes that structure is itself emergent and changes along with the agents inside it, so the way to intervene is to shape conditions, run probes and amplify what emerges.

The Prairie Oyster essay puts the diagnostic question well: is this “a system you can design, or one you can only navigate”?

Integration is already happening

Jackson (2019) reorganises critical systems thinking around complexity, and argues for choosing a method according to the type of complexity faced. Snowden and Rancati (2021) take systems mapping in as one tool among many, suited to complicated rather than complex situations.

So the practical conclusion is not to abandon systems thinking. It is to know where it is valid. A loop diagram of how homework is set and marked is useful. It stops being useful the week parents start paying for tuition because of what the ratings seem to say.

Misreadings

What complexity thinking is not

Not chaos

Complex systems are often highly ordered. The order is just not central. Nobody at the school is in charge of which subjects students think are harsh, and there is still a ranking, stable enough to read.

Not “everything is connected”

Near-decomposability means most things are weakly connected most of the time (Simon, 1962). The skill is finding the few strong couplings, not drawing a line from everything to everything.

Not anti-planning

It changes what a plan is for. A plan stops specifying outcomes and starts setting direction and constraints. There is still a plan, and someone still owns it.

Not relativism

Some models are better than others, even though none is complete. Holding several models is a way of finding out which conclusions survive. It is not permission to believe whichever one suits.

Failure modes

Eight ways it goes wrong, each with a guardrail

Complexity thinking has to meet its own standards: provisional, plural, reflexive and testable. A complexity practice that cannot say what would show it to be wrong has become the thing it criticises.

Failure modes, what each looks like, and its guardrail
Failure	What it looks like	Guardrail
Complexity as alibi	“It’s complex” used to avoid commitment or accountability.	Commit to a direction, learning milestones and explicit triggers.
Paralysis by optionality	Options kept open indefinitely. Nothing is decided.	Use the optionality grid, and time-box each stage.
Category error	Complexity methods applied to clear or complicated problems.	Apply the three-condition test and the six dials first.
Metaphor creep	Concepts used as metaphors without mechanisms.	Demand mechanisms and evidence. Keep complexity science and complexity thinking distinct.
Romantic holism	“Everything is connected”, so nothing can be prioritised.	Near-decomposability: identify the strong couplings and act on them.
Developmental elitism	Complexity thinking cast as a higher stage held by an enlightened few.	Treat the capacities as distributed and teachable. Rely on collective intelligence.
Crisis atmospherics	Crisis language used to create urgency without naming a mechanism.	Name the crisis type, its mechanism and the decision it calls for.
Warm-data vagueness	Rich conversation with no traceable effect.	Pair it with cold data and probes. Record what changed as a result.

Reading

Where it comes from

Start with the first on the list.

Ladyman, J., Lambert, J., & Wiesner, K. (2013). What is a complex system? European Journal for Philosophy of Science, 3(1), 33–67.
The careful account of why feature lists describe complex systems without saying when ordinary reasoning breaks.
Anderson, P. W. (1972). More is different. Science, 177(4047), 393–396.
The short paper behind the first condition: wholes that cannot be read off their parts.
Holland, J. H. (1995). Hidden order: How adaptation builds complexity. Addison-Wesley.
The source for the second condition, on agents and rules that change as the system runs.
Stacey, R. D. (2001). Complex responsive processes in organizations: Learning and knowledge creation. Routledge.
The strongest case that the manager is inside the pattern, and the root of the third condition.
Heifetz, R. A. (1994). Leadership without easy answers. Belknap Press of Harvard University Press.
The distinction between technical problems and adaptive challenges that marks what the test excludes.
Kurtz, C. F., & Snowden, D. J. (2003). The new dynamics of strategy: Sense-making in a complex and complicated world. IBM Systems Journal, 42(3), 462–483.
Cynefin’s original argument that the first job is to sort situations by type, still the most underused idea in the field.
Cilliers, P. (1998). Complexity and postmodernism: Understanding complex systems. Routledge.
Why every model of a complex system leaves something out, and why that makes framing an ethical act.
Simon, H. A. (1962). The architecture of complexity. Proceedings of the American Philosophical Society, 106(6), 467–482.
Near-decomposability, the answer to anyone who says everything is connected.
Jackson, M. C. (2019). Critical systems thinking and the management of complexity. Wiley.
The best account of how systems thinking and complexity thinking are already being joined up.
Byrne, D., & Callaghan, G. (2014). Complexity theory and the social sciences: The state of the art. Routledge.
A sober guide to the gap between complexity science and complexity used loosely in social settings.

Unsettled

What I would still argue about

The questions this page does not settle.

How would anyone show that the three-condition test sorts situations better than experienced judgement does? The frameworks it builds on have far more case illustrations than controlled evaluations.
Most real situations are mixed: a technical core inside an adaptive challenge, like the tracking system inside the assessment plan. It is not settled whether the test should be applied to the whole situation or to its parts, or who decides where the parts divide.
The test says when complexity thinking is needed. It says nothing about which outcomes are worth wanting. Complexity science is thin on values, and the attempts to supply them have had uneven success.
Committing to direction and learning instead of fixed outputs is the answer to complexity as alibi. It can also be a softer way of avoiding commitment. The line between the two is easier to state than to hold in public.
The canon behind the three conditions is mostly Western and institutionally narrow. Whether relational traditions would draw the conditions the same way, or draw different ones, is an open question.

The seven contributions

Singapore